This is the README file for the BOLD Moments Dataset ./derivatives/stimuli_metadata folder. Contents of this directory correspond to derivatives of the stimulus set, such as features, annotations, and frames. ME_feats_matlab: Contains motion energy (ME) features of each video computed from the Matlab implementation of the motion energy model (https://github.com/gallantlab/motion_energy_matlab). annotations.json: metadata for the stimulus set. annotations_filednames.json: Descriptions of metadata fields corresponding to the 'annotations.json' folder. llm_frame_annotations.json: Contains five sentence text descriptions of just the middle frame of each video (from 'frames_middle' folder) generated by GIT model version git-large-coco (Wang et al., 2022). Due to copyright permission considerations, you must separately download the actual .mp4 stimulus videos and frames (if you agree to the terms of use). Go to this link [https://boldmomentsdataset.csail.mit.edu/stimuli_metadata] to download the stimulus set files and agree to the terms of use: Inside that stimulus set folder (and if you agree to the terms of use), you will see: frames: Contains each stimulus video's frames as a .jpg file. Useful for model inputs. stimuli: The experimental stimulus set that otherwise would be in the root directory for a BIDS compatible folder structure. ./stimuli/test/ and ./stimuli/train/ folders contain the test and train stimuli, respectively, and the ./stimuli/localizer/ folder contains the stimulus set used for the functional localizer task. mp4_h264: The stimuli set reformatted into H.264 compressed .mp4 files. Mainly reformatted for convenience, this format is slightly more flexible across platforms for loading and viewing, such as if you wish to load the videos in a web browser for additional annotation experiments. frames_middle: Contains a .jpg file of the middle frame of each video. This middle frame was also used as model input for the frame annotations in "llm_frame_annotations.json". stimuli_access.txt: text file containing the terms of use and password to access the dataset.